Déborah MOERLEN – Modeling of an ETDR sensor for distributed high-temperature measurement (>1000°C): experimental validations, durability and Machine Learning

Jury

Dr. Guylaine Poulin-Vittrant, Research Director at CNRS Centre Limousin Poitou Charente, Reviewer
Prof. Lionel Duvillaret, Professor and President of Kaptéos SAS, Reviewer
Prof. Alain Sylvestre, Professor at Université Grenoble Alpes, Chair of the Jury
Dr. Virgil Optasanu, Associate Professor at Université de Bourgogne, Examiner
Prof. Fabien Volpi, Professor at Université Grenoble Alpes, Supervisor of the thesis
Dr. Valérie Parry, Associate Professor at Université Grenoble Alpes, Co-Supervisor of the thesis
Dr. Mickaël Boinet, R&D Engineer at Saint-Gobain Research Provence, Invited Member
Mr. Jean-Luc Parouty, Research Engineer at CNRS, SIMaP Laboratory, Invited Member
 

Abstract

Real-time control of structural deviations in industrial installations operating at high temperatures (> 1000°C) is a critical safety concern. Integrated Structural Health Monitoring (SHM) applied to these installations enables both the correction of process deviations on production lines and intervention before failures occur. For temperature measurement, thermocouples can provide local measurements until very high temperatures, but they cannot monitor potential deviations at any point of an installation. An alternative, Optical Time-Domain Reflectometry (OTDR), has demonstrated its capability to measure distributed temperatures. However, optical fibers cannot withstand temperatures above 600°C for extended durations. To address this dual limitation, this PhD work contributed to developing a sensor capable of continuously measuring and spatially mapping temperatures exceeding 1000°C in industrial settings. The sensor under study utilizes the Electrical Time-Domain Reflectometry (ETDR) technique, which involves analyzing the high-frequency electromagnetic signal backscattered by a transmission line. A proof of concept was conducted using a bifilar transmission line made of refractory materials. Preliminary analyses confirmed the feasibility of extracting a temperature profile from the signals backscattered by heterogeneities along the line. However, to enhance sensitivity and durability, it is essential to quantify the impact of the sensor’s characteristics: its geometry, the physico-chemical properties of its materials, and their evolutions over time and temperature. Therefore, three multidisciplinary research axes were explored: i) identifying the parameters influencing the experimental reflectograms and quantifying their effects, ii) optimizing the sensor, and iii) proposing new approaches for signal processing. The ETDR sensor was initially modeled, based on its geometry, material properties (permittivity, resistivity, permeability, thermal expansion, etc.), and their temperature dependencies (experimentally measured). This primarily analytical model was supplemented by finite element calculations. Without any adjustable parameters, all experimental observations were replicated, particularly the sensitivity to temperature variations. A parametric study was then conducted, identifying an optimal design for this bifilar sensor. To select the most efficient materials for this sensor, durability tests were performed, correlating the evolution of backscattered signals with material aging. This approach identified platinum FKS and Kanthal AF © as the optimal conductive materials. However, a surface protection solution may be developed to limit their aging under real conditions. Finally, a machine learning-based method for reflectogram analysis was developed to extract temperature profiles from simulated sensors. This method showed very promising results for obtaining highly sensitive profiles. It could thus enhance existing signal processing based on cross-correlation between reflectograms. In conclusion, this study demonstrates the feasibility of continuously and spatially measuring temperatures exceeding 1000°C with this type of sensor. These advancements have significant implications for improving the monitoring of high-temperature industrial installations, potentially leading to more reliable and efficient structural monitoring solutions.